Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Tue, 18 May 2010 12:23:41 -0500

Hi John,

On Monday 17 May 2010 14:09:54 John Dobbins wrote:
> question 1) does Channel Access transfer only NELM number of values or is
> it transferring MALM number of values?

CA transfers the number of values requested, and will zero-fill any values 
beyond the number of elements that the record type reports; caget lets you 
limit the request size, but it defaults to the whole array as reported by the 
record type's cvt_dbaddr() routine at connection time.

> question 2) would this break down if I attach a monitor to my_subArray?
> i.e. will changing NELM cause problems for a monitor?

No, but IIRC CA currently fixes the amount of data that will be sent at 
subscription time; if the client specifies zero for the number of elements it 
will be taken as the array size reported by the record type's cvt_dbaddr() 
routine at connection time.

> question 3) does aSub field NEA (number of input A elements) get set by db
> processing when input A is read in? i.e. can my subroutine use the value of
> NEA to tell how many values were read in?

NEA does get updated when the input link is read *provided* that it's a DB 
link.  CA links don't know how much data was actually transferred, so NEA will 
always be set to NOA in that case, but for DB links the size should be 
correctly set afterwards.

> question 4) is my aSub subroutine allowed to change NEVA (number of input A
> elements) (Record Reference manual field summary says "no modify" - or does
> that refer only to via dbPut?)

That summary is only for external access to the field, your subroutine is 
allowed to change NEVA.

> question 5a) if answer to 4 is yes, will "caget my_aSub.VALA" return NEVA
> number of elements?

No, it will return NOVA elements, but only the first NEVA will come from VALA, 
the remaining elements will be zero.  CA doesn't tell you what value NEVA had 
at the time though, and unfortunately monitors do not get posted on the NEVA 
field so you'd have to do an explicit ca_get() to find it.

You might be able to utilize the first array element to hold the size if this 
is important to you and you have control over the client code that will be 
accessing the array.  This is a limitation of the current CA implementation 
and API.

HTH,

- Andrew
-- 
The best FOSS code is written to be read by other humans -- Harald Welte

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System